MatX is developing MatX One, an AI accelerator that keeps model weights in SRAM for high-throughput, low-latency inference on large models in very large clusters.
Ex-Google TPU founders building an LLM-specific chip (SRAM-first with HBM) at TSMC for 2027 shipment; $500M Series B from Jane Street and Situational Awareness with Marvell as strategic investor.
6 signals name MatX. Members see all of them, filtered by topic, source and date.
10 facts on file. Members see all of them with the source link.
Traction: deployment, production, commercials, progress. Capability: technology, positioning, firepower, leverage.
MatX is developing MatX One, an AI accelerator that keeps model weights in SRAM for high-throughput, low-latency inference on large models in very large clusters. Ex-Google TPU founders building an LLM-specific chip (SRAM-first with HBM) at TSMC for 2027 shipment; $500M Series B from Jane Street and Situational Awareness with Marvell as strategic investor.
MatX is on the AI Core watchlist list of Frontier Tech Intelligence with a score of 54 out of 100, below the Top 100.
MatX is based in United States.
Subscribers see the full file, every signal, the live ranking and the questions the signals answer: where money moved, who controls the layer, where supply is constrained. $99 a month, 7-day trial. What Frontier Tech Intelligence is →